A method and system for large-pose face alignment based on 3D models

Abstract

The invention discloses a large-pose face alignment method based on a 3D model, including: establishing a PCA-based 3DMM model, constructing a vector set U composed of N two-dimensional facial feature points collected by a 3D facial scanner, and constructing The relationship between the 3DMM model and the constructed vector set U is used to construct an improved CNN network model, which is implemented by adding a visualization layer to each visualization block in the existing CNN network model, which is used to convert the features of the visualization block where it is located Extract it and pass it to the next visualization block, obtain training samples based on the existing public face data set, use the training samples to train the improved CNN network model to obtain corresponding parameters, and obtain two-dimensional face pictures , and input the two-dimensional face picture into the trained improved CNN network model. The invention can solve the technical problem that the existing large-pose human face alignment method cannot take into account the density of feature points and the convergence speed in the training stage at the same time.

Description

technical field [0001] The invention belongs to the technical field of pattern recognition, and more specifically, relates to a method and system for aligning large-pose human faces based on a 3D model. Background technique [0002] At present, face alignment technology has become a new research hotspot in the field of computer vision, which is the process of aligning facial elements (such as eyes, nose, mouth, outline). Accurate face alignment is an important prerequisite for many face-related tasks, including face recognition, 3D face reconstruction, face animation, etc. [0003] In the research of face alignment technology, large pose (ie ±90° face angle) face alignment (Large poseface alignment, LPFA for short) is a research difficulty. At present, the sparse feature point estimation method is usually used, but this will greatly reduce the effect of face alignment (for example, in the case of side faces, the face alignment effect will be poor). In order to solve this p...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the above defects or improvement needs of the prior art, the present invention provides a method and system for large-pose face alignment based on 3D models, the purpose of which is to achieve large-pose face alignment by using dense feature point estimation while , to ensure high training and convergence speed, so as to solve the technical problem that existing large-pose face alignment methods cannot take into account both the density of feature points and the convergence speed in the training phase

Images